Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'dev' into patch-1

Browse Source
This commit is contained in:
Vladimir Mandic
2025-11-04 07:23:20 -05:00
committed by GitHub
parent 95352051a9 a4378a79e4
commit 9044aca2c7
1 changed files with 21 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sdnq/common.py
+21 -14
View File
@@ -83,10 +83,7 @@ else:
common_skip_keys = (
".lm_head",
".patch_embedding",
".time_embed",
".time_text_embed",
".context_embedder",
".condition_embedder",
".x_embedder",
@@ -101,36 +98,45 @@ common_skip_keys = (
".img_out",
".vid_out",
".final_layer",
"pos_embed",
"time_text_embed",
"patch_embedding",
"patch_embed",
"patch_emb",
"lm_head",
"wte",
)
module_skip_keys_dict = {
"FluxTransformer2DModel": [
["single_transformer_blocks.0.norm.linear.weight", ".time_text_embed", ".context_embedder", ".x_embedder", ".proj_out", ".norm_out", "pos_embed"],
["single_transformer_blocks.0.norm.linear.weight", "time_text_embed", "context_embedder", "x_embedder", ".proj_out", "norm_out"],
{}
],
"ChromaTransformer2DModel": [
["distilled_guidance_layer", ".time_text_embed", ".context_embedder", ".x_embedder", ".proj_out", ".norm_out", "pos_embed"],
["distilled_guidance_layer", "time_text_embed", "context_embedder", "x_embedder", ".proj_out", "norm_out"],
{}
],
"QwenImageTransformer2DModel": [
["transformer_blocks.0.img_mod.1.weight", ".time_text_embed", ".txt_in", ".img_in", ".proj_out", ".norm_out", "pos_embed"],
["transformer_blocks.0.img_mod.1.weight", "time_text_embed", "txt_in", "img_in", "proj_out", "norm_out"],
{}
],
"WanTransformer3DModel": [
["scale_shift_table", ".rope", ".patch_embedding", ".condition_embedder", ".proj_out", ".norm_out", "pos_embed"],
{}
],
"Gemma3nForCausalLM": [
[".lm_head", "correction_coefs", "prediction_coefs", "embedding_projection"],
["scale_shift_table", "patch_embedding", "condition_embedder", "proj_out", "norm_out"],
{}
],
"HunyuanImage3ForCausalMM": [
[".lm_head", ".patch_embed", ".time_embed", ".time_embed_2", ".final_layer", ".model.wte", ".model.ln_f", ".timestep_emb", ".vae", ".vision_aligner", ".vision_model.head", ".vision_model.post_layernorm", ".vision_model.embeddings"],
["lm_head", "patch_embed", "time_embed", "time_embed_2", "final_layer", "wte", "ln_f", "timestep_emb", "vae", "vision_aligner", "head", "post_layernorm", "embeddings"],
{}
],
"Emu3ForCausalLM": [
[".lm_head", ".vq_model", ".tokenizer"],
["lm_head", "vq_model", "tokenizer"],
{}
],
"Gemma3nForCausalLM": [
["lm_head", "correction_coefs", "prediction_coefs", "embedding_projection"],
{}
],
"MoondreamModel": [
["lm_head", "region", "wte", "post_ln", "proj_mlp", "patch_emb", "pos_emb"],
{}
],
"NaDiT": [
@@ -141,4 +147,5 @@ module_skip_keys_dict = {
module_skip_keys_dict["ChronoEditTransformer3DModel"] = module_skip_keys_dict["WanTransformer3DModel"]
module_skip_keys_dict["Gemma3nForConditionalGeneration"] = module_skip_keys_dict["Gemma3nForCausalLM"]
module_skip_keys_dict["HfMoondream"] = module_skip_keys_dict["MoondreamModel"]
module_skip_keys_dict["NaDiTUpscaler"] = module_skip_keys_dict["NaDiT"]